Search

Unit tests represent the most basic level of testing within the software testing lifecycle and are crucial to ensuring software correctness. Designing and creating unit tests is a costly and labor-intensive process that is ripe for automation. Recently, Large Language Models (LLMs) have been applied to various aspects of software development, including unit test generation. Although several empirical studies evaluating LLMs' capabilities in test code generation exist, they primarily focus on simple scenarios, such as the straightforward generation of unit tests for individual methods. These evaluations often involve independent and small-scale test units, providing a limited view of LLMs' performance in real-world software development scenarios. Moreover, previous studies do not approach the problem at a suitable scale for real-life applications. Generated unit tests are often evaluated via manual integration into the original projects, a process that limits the number of tests executed and reduces overall efficiency. To address these gaps, we have developed an approach for generating and evaluating more real-life complexity test suites. Our approach focuses on class-level test code generation and automates the entire process from test generation to test assessment. In this work, we present AgoneTest: an automated system for generating test suites for Java projects and a comprehensive and principled methodology for evaluating the generated test suites. Starting from a state-of-the-art dataset (i.e., Methods2Test), we built a new dataset for comparing human-written tests with those generated by LLMs. Our key contributions include a scalable automated software system, a new dataset, and a detailed methodology for evaluating test quality.

Context: Molecular hydrogen ($\rm{H_2}$) is crucial in galaxy formation and evolution, serving as the main fuel for star formation (SF). In metal-enriched environments, $\rm{H_2}$ primarily forms on interstellar dust grain surfaces. However, due to the complexities of modelling this process, SF in cosmological simulations often relies on empirical or theoretical frameworks validated only in the Local Universe to estimate the abundance of $\rm{H_2}$. Aims: This study aims to model the connection between star, dust, and $\rm{H_2}$ formation processes in cosmological simulations. Methods: We include $\rm{H_2}$ formation on dust grain surfaces and account for molecule destruction and radiation shielding into the SF and feedback model MUPPI. Results: The model reproduces key properties of observed galaxies for stellar, dust, and $\rm{H_2}$ components. The cosmic density of $\rm{H_2}$ ($\rho_{\rm{H2}}$) peaks around $z=1.5$, then decreases by half towards $z=0$, showing milder evolution than observed. The $\rm{H_2}$ mass function since $z=2$ also shows gentler evolution. Our model successfully recovers the integrated molecular Kennicutt-Schmidt (mKS) law between surface star formation rate ($\Sigma_{\rm SFR}$) and surface $\rm{H_2}$ density ($\Sigma_{\rm H2}$) at $z=0$, already evident at $z=2$ with a higher normalization. We find hints of a broken power law with a steeper slope at higher $\Sigma_{\rm H2}$, aligning with some observational findings. Additionally, the $\rm{H_2}$-to-dust mass ratio in galaxies shows a decreasing trend with gas metallicity and stellar mass. The $\rm{H_2}$-to-dust mass fraction for the global galaxy population is higher at higher redshifts. The analysis of the atomic-to-molecular transition on a particle-by-particle basis suggests that gas metallicity cannot reliably substitute the dust-to-gas ratio in models simulating dust-promoted $\rm{H_2}$., Comment: submitted to A&A. Comments are welcome

Recently, there has been growing attention on behalf of both academic and practice communities towards the ability of Artificial Intelligence (AI) systems to operate responsibly and ethically. As a result, a plethora of frameworks and guidelines have appeared to support practitioners in implementing Trustworthy AI applications (TAI). However, little research has been done to investigate whether such frameworks are being used and how. In this work, we study the vision AI practitioners have on TAI principles, how they address them, and what they would like to have - in terms of tools, knowledge, or guidelines - when they attempt to incorporate such principles into the systems they develop. Through a survey and semi-structured interviews, we systematically investigated practitioners' challenges and needs in developing TAI systems. Based on these practical findings, we highlight recommendations to help AI practitioners develop Trustworthy AI applications.

This workshop proposal focuses on best practices in UI/UX design for AI applications aimed at children, emphasising safety, engagement, and ethics. It aims to address the challenge of measuring the safety, trustworthiness, and reliability of interactions between children and AI systems. Through collaborative discussions, participants will explore effective design strategies and ethical guidelines while developing methodologies for assessing the safety and reliability of AI interactions with children. This proposal seeks to foster responsible and child-centered AI design practices within the CHI community., Comment: 4 pages, The paper has been peer-reviewed and presented at the "CHI 2024 Workshop on Child-centred AI Design", May 11, 2024, Honolulu, HI, USA

We implement a sub-resolution prescription for the unresolved dynamical friction onto black holes (BHs) in the OpenGadget3 code. We carry out cosmological simulations of a volume of 16 cMpc3 and zoom-ins of a galaxy group and of a galaxy cluster. The advantages of our new technique are assessed in comparison to commonly adopted methods to hamper spurious BH displacements, i.e. repositioning onto a local minimum of the gravitational potential and ad-hoc boosting of the BH particle dynamical mass. The newly-introduced dynamical friction correction provides centering of BHs on host halos which is at least comparable with the other techniques. It predicts half as many merger events with respect to the repositioning prescription, with the advantage of being less prone to leave sub-structures without any central BH. Simulations featuring our dynamical friction prescription produce a smaller (by up to 50% with respect to repositioning) population of wandering BHs and final BH masses in good agreement with observations. As for individual BH-BH interactions, our dynamical friction model captures the gradual inspiraling of orbits before the merger occurs. By contrast, the repositioning scheme, in its most classical renditions considered, describes extremely fast mergers, while the dynamical mass misrepresents the BHs' dynamics, introducing numerical scattering between the orbiting BHs. Given its performances in describing the centering of BHs within host galaxies and the orbiting of BH pair before their merging, our dynamical friction correction opens interesting applications for an accurate description of the evolution of BH demography within cosmological simulations of galaxy formation at different cosmic epochs and within different environments., Comment: 19 pages, 10 figures, submitted to A&A

This thesis studies a pair of symmetry protected topological (SPT) phases which arise when considering one-dimensional quantum spin systems possessing a natural orthogonal group symmetry. Particular attention is given to a family of exactly solvable models whose ground states admit a matrix product state description and generalize the AKLT chain. We call these models ``$SO(n)$ AKLT chains'' and the phase they occupy the ``$SO(n)$ Haldane phase''. We present new results describing their ground state structure and, when $n$ is even, their peculiar $O(n)$-to-$SO(n)$ symmetry breaking. We also prove that these states have arbitrarily large correlation and injectivity length by increasing $n$, but all have a 2-local parent Hamiltonian, in contrast to the natural expectation that the interaction range of a parent Hamiltonian should diverge as these quantities diverge. We extend Ogata's definition of an SPT index for a split state for a finite symmetry group $G$ to an SPT index for a compact Lie group $G$. We then compute this index, which takes values in the second Borel group cohomology $H^2(SO(n),U(1))$, at a single point in each of the SPT phases. The two points have different indices, confirming the two SPT phases are indeed distinct. Chapter 1 contains an introduction with a detailed overview of the contents of this thesis, which includes several chapters of background information before presenting new results in Chapter 7 and Chapter 8., Comment: Dissertation for a PhD in Mathematics at UC Davis

In recent months, the social impact of Artificial Intelligence (AI) has gained considerable public interest, driven by the emergence of Generative AI models, ChatGPT in particular. The rapid development of these models has sparked heated discussions regarding their benefits, limitations, and associated risks. Generative models hold immense promise across multiple domains, such as healthcare, finance, and education, to cite a few, presenting diverse practical applications. Nevertheless, concerns about potential adverse effects have elicited divergent perspectives, ranging from privacy risks to escalating social inequality. This paper adopts a methodology to delve into the societal implications of Generative AI tools, focusing primarily on the case of ChatGPT. It evaluates the potential impact on several social sectors and illustrates the findings of a comprehensive literature review of both positive and negative effects, emerging trends, and areas of opportunity of Generative AI models. This analysis aims to facilitate an in-depth discussion by providing insights that can inspire policy, regulation, and responsible development practices to foster a human-centered AI., Comment: Presented at GoodIT2023 - ACM Conference on Information Technology for Social Good

In the ever-expanding landscape of Artificial Intelligence (AI), where innovation thrives and new products and services are continuously being delivered, ensuring that AI systems are designed and developed responsibly throughout their entire lifecycle is crucial. To this end, several AI ethics principles and guidelines have been issued to which AI systems should conform. Nevertheless, relying solely on high-level AI ethics principles is far from sufficient to ensure the responsible engineering of AI systems. In this field, AI professionals often navigate by sight. Indeed, while recommendations promoting Trustworthy AI (TAI) exist, these are often high-level statements that are difficult to translate into concrete implementation strategies. There is a significant gap between high-level AI ethics principles and low-level concrete practices for AI professionals. To address this challenge, our work presents an experience report where we develop a novel holistic framework for Trustworthy AI - designed to bridge the gap between theory and practice - and report insights from its application in an industrial case study. The framework is built on the result of a systematic review of the state of the practice, a survey, and think-aloud interviews with 34 AI practitioners. The framework, unlike most of those already in the literature, is designed to provide actionable guidelines and tools to support different types of stakeholders throughout the entire Software Development Life Cycle (SDLC). Our goal is to empower AI professionals to confidently navigate the ethical dimensions of TAI through practical insights, ensuring that the vast potential of AI is exploited responsibly for the benefit of society as a whole.

Mobile robotic systems are becoming increasingly popular. These systems are used in various indoor applications, raging from warehousing and manufacturing to test benches for assessment of advanced control strategies, such as artificial intelligence (AI)-based control solutions, just to name a few. Scaled robotic cars are commonly equipped with a hierarchical control acthiecture that includes tasks dedicated to vehicle state estimation and control. This paper covers both aspects by proposing (i) a federeted extended Kalman filter (FEKF), and (ii) a novel deep reinforcement learning (DRL) path tracking controller trained via an expert demonstrator to expedite the learning phase and increase robustess to the simulation-to-reality gap. The paper also presents the formulation of a vehicle model along with an effective yet simple procedure for identifying tis paramters. The experimentally validated model is used for (i) supporting the design of the FEKF and (ii) serving as a digital twin for training the proposed DRL-based path tracking algorithm. Experimental results confirm the ability of the FEKF to improve the estimate of the mobile robot's position. Furthermore, the effectiveness of the DRL path tracking strateguy is experimentally tested along manoeuvres not considered during training, showing also the ability of the AI-based solution to outpeform model-based control strategies and the demonstrator. The comparison with benchmraking controllers is quantitavely evalueted through a set of key performance indicators., Comment: Under review on IEEE Transactions

The large-scale environment of the cosmic web is believed to impact galaxy evolution, but there is still no consensus regarding the mechanisms. We use a semi-analytic model (SAM) galaxy catalog to study the star formation and dust content of local galaxies in different cosmic environments of the cosmic web, namely voids, filaments, walls, and nodes. We find a strong impact of the environment only for galaxies with $M_{\rm stars}\lesssim10^{10.8}\, M_\odot$: the less dense the environment, the larger the star formation rate and dust content at fixed stellar mass. This is attributed to the fact that galaxies in less dense environments typically feature younger stellar populations, a slower evolution of their stellar mass and a delayed star formation compared to galaxies in denser environments. As for galaxies with $M_{\rm stars}\gtrsim 10^{10.8}\, M_\odot$ differences among environments are milder due to the disc instability (DI) driven supermassive black hole (SMBH) growth implemented in the SAM, which makes SMBH growth, and thus galaxy quenching, environment insensitive. We qualitatively test our predictions against observations by identifying environments in the SDSS-DR16 using dust masses derived from the GAMA survey. The agreement is encouraging, particularly at ${\rm log} \, M_{\rm stars}/M_\odot\gtrsim 10.5-11$, where sSFRs and dust masses appear quite environment-insensitive. This result confirms the importance of in situ growth channels of SMBHs., Comment: 21 pages, 12 figures, 3 tables. Accepted for publication in ApJ

Variational quantum computing schemes train a loss function by sending an initial state through a parametrized quantum circuit, and measuring the expectation value of some operator. Despite their promise, the trainability of these algorithms is hindered by barren plateaus (BPs) induced by the expressiveness of the circuit, the entanglement of the input data, the locality of the observable, or the presence of noise. Up to this point, these sources of BPs have been regarded as independent. In this work, we present a general Lie algebraic theory that provides an exact expression for the variance of the loss function of sufficiently deep parametrized quantum circuits, even in the presence of certain noise models. Our results allow us to understand under one framework all aforementioned sources of BPs. This theoretical leap resolves a standing conjecture about a connection between loss concentration and the dimension of the Lie algebra of the circuit's generators., Comment: 11+17 pages, 2+2 figures, updated manuscript and title to published version

Various studies identified possible drivers of extremes of Arctic sea ice reduction, such as observed in the summers of 2007 and 2012, including preconditioning, local feedback mechanisms, oceanic heat transport and the synoptic- and large-scale atmospheric circulations. However, a robust quantitative statistical analysis of extremes of sea ice reduction is hindered by the small number of events that can be sampled in observations and numerical simulations with computationally expensive climate models. Recent studies tackled the problem of sampling climate extremes by using rare event algorithms, i.e., computational techniques developed in statistical physics to reduce the computational cost required to sample rare events in numerical simulations. Here we apply a rare event algorithm to ensemble simulations with the intermediate complexity coupled climate model PlaSim-LSG to investigate extreme negative summer pan-Arctic sea ice area anomalies under pre-industrial greenhouse gas conditions. Owing to the algorithm, we estimate return times of extremes orders of magnitude larger than feasible with direct sampling, and we compute statistically significant composite maps of dynamical quantities conditional on the occurrence of these extremes. We find that extremely low sea ice summers in PlaSim-LSG are associated with preconditioning through the winter sea ice-ocean state, with enhanced downward longwave radiation due to an anomalously moist and warm spring Arctic atmosphere and with enhanced downward sensible heat fluxes during the spring-summer transition. As a consequence of these three processes, the sea ice-albedo feedback becomes active in spring and leads to an amplification of pre-existing sea ice area anomalies during summer.

We investigate the statistics and dynamics of extreme heat waves over different areas of Europe. We find heatwaves over France and Scandinavia to be associated with recurrent wavenumber three teleconnection patterns in surface temperature and mid-tropospheric geopotential height. For heatwaves with return times of 4 years these teleconnection patterns and their dynamics are robustly represented in a hierarchy of models of different complexity and in reanalysis data. For longer return times, reanalysis records are too short to give statistically significant results, while models confirm the relevance of these large scale patterns for the most extreme heatwaves. A time series analysis shows that heatwave indices defined at synoptic scale are fairly well described by Gaussian stochastic processes, and that these Gaussian processes reproduce well return time plots even for very rare events. These results suggest that extreme heatwaves over different areas of Europe show recurrent typical behaviours in terms of long-range spatial correlations and subseasonal-scale temporal correlations. These properties are consistently represented among models of different complexity and observations, thus suggesting their relevance for a better understanding of the drivers and causes of the occurrence of extreme midlatitude heatwaves and their predictability., Comment: 25 pages, 11 figures

In the last years, the raise of Artificial Intelligence (AI), and its pervasiveness in our lives, has sparked a flourishing debate about the ethical principles that should lead its implementation and use in society. Driven by these concerns, we conduct a rapid review of several frameworks providing principles, guidelines, and/or tools to help practitioners in the development and deployment of Responsible AI (RAI) applications. We map each framework w.r.t. the different Software Development Life Cycle (SDLC) phases discovering that most of these frameworks fall just in the Requirements Elicitation phase, leaving the other phases uncovered. Very few of these frameworks offer supporting tools for practitioners, and they are mainly provided by private companies. Our results reveal that there is not a "catching-all" framework supporting both technical and non-technical stakeholders in the implementation of real-world projects. Our findings highlight the lack of a comprehensive framework encompassing all RAI principles and all (SDLC) phases that could be navigated by users with different skill sets and with different goals.

Current AI regulations require discarding sensitive features (e.g., gender, race, religion) in the algorithm's decision-making process to prevent unfair outcomes. However, even without sensitive features in the training set, algorithms can persist in discrimination. Indeed, when sensitive features are omitted (fairness under unawareness), they could be inferred through non-linear relations with the so called proxy features. In this work, we propose a way to reveal the potential hidden bias of a machine learning model that can persist even when sensitive features are discarded. This study shows that it is possible to unveil whether the black-box predictor is still biased by exploiting counterfactual reasoning. In detail, when the predictor provides a negative classification outcome, our approach first builds counterfactual examples for a discriminated user category to obtain a positive outcome. Then, the same counterfactual samples feed an external classifier (that targets a sensitive feature) that reveals whether the modifications to the user characteristics needed for a positive outcome moved the individual to the non-discriminated group. When this occurs, it could be a warning sign for discriminatory behavior in the decision process. Furthermore, we leverage the deviation of counterfactuals from the original sample to determine which features are proxies of specific sensitive information. Our experiments show that, even if the model is trained without sensitive features, it often suffers discriminatory biases.

The increasing application of Artificial Intelligence and Machine Learning models poses potential risks of unfair behavior and, in light of recent regulations, has attracted the attention of the research community. Several researchers focused on seeking new fairness definitions or developing approaches to identify biased predictions. However, none try to exploit the counterfactual space to this aim. In that direction, the methodology proposed in this work aims to unveil unfair model behaviors using counterfactual reasoning in the case of fairness under unawareness setting. A counterfactual version of equal opportunity named counterfactual fair opportunity is defined and two novel metrics that analyze the sensitive information of counterfactual samples are introduced. Experimental results on three different datasets show the efficacy of our methodologies and our metrics, disclosing the unfair behavior of classic machine learning and debiasing models.

Observations suggest that the amount of galactic dust in the Universe decreased by a factor $\sim 2-3$ during the last $\sim 8$ Gyr. However, cosmological models of galaxy evolution usually struggle to explain this decrease. Here we use the semi-analytic model (SAM) L-Galaxies2020 to show that this drop may be reproduced assuming standard prescriptions for dust production and evolution. We extend the SAM with i) a state-of-the-art dust model which adopts the two-size approximation and ii) a new disc instability criterion which triggers bulge and central black hole growth. The model reproduces some fundamental properties of the local galaxy population, such as the fraction of spheroid-dominated galaxies and some scaling relations involving dust. Moreover, the model predicts a galactic dust drop from $z \sim 1 \rightarrow 0$, which becomes closer to the observed one when adopting the new treatment of disc instabilities. This result is related to the newly implemented super-massive black hole growth during disc instabilities, which enhances the quenching of massive galaxies. Consequently, these objects feature a lower gas and dust content. We provide a census of the contribution of all the processes affecting the galactic dust content. Accretion is the dominant dust mass growth process. Destruction by supernovae, astration and ejection by winds have all a non-negligible role in decreasing the overall dust content in galaxies below $z \sim 1$. We also discuss predictions concerning extra-galactic dust, confirming that a sputtering efficiency lower than the canonical one is required to match the few available observations., Comment: 20 pages, 20 figures, 1 table. Accepted for publication in MNRAS

Quantum neural network architectures that have little-to-no inductive biases are known to face trainability and generalization issues. Inspired by a similar problem, recent breakthroughs in machine learning address this challenge by creating models encoding the symmetries of the learning task. This is materialized through the usage of equivariant neural networks whose action commutes with that of the symmetry. In this work, we import these ideas to the quantum realm by presenting a comprehensive theoretical framework to design equivariant quantum neural networks (EQNN) for essentially any relevant symmetry group. We develop multiple methods to construct equivariant layers for EQNNs and analyze their advantages and drawbacks. Our methods can find unitary or general equivariant quantum channels efficiently even when the symmetry group is exponentially large or continuous. As a special implementation, we show how standard quantum convolutional neural networks (QCNN) can be generalized to group-equivariant QCNNs where both the convolution and pooling layers are equivariant to the symmetry group. We then numerically demonstrate the effectiveness of a SU(2)-equivariant QCNN over symmetry-agnostic QCNN on a classification task of phases of matter in the bond-alternating Heisenberg model. Our framework can be readily applied to virtually all areas of quantum machine learning. Lastly, we discuss about how symmetry-informed models such as EQNNs provide hopes to alleviate central challenges such as barren plateaus, poor local minima, and sample complexity., Comment: 26+21 pages, 12 + 2 figures; journal version with new numerics section

Recent advances in classical machine learning have shown that creating models with inductive biases encoding the symmetries of a problem can greatly improve performance. Importation of these ideas, combined with an existing rich body of work at the nexus of quantum theory and symmetry, has given rise to the field of Geometric Quantum Machine Learning (GQML). Following the success of its classical counterpart, it is reasonable to expect that GQML will play a crucial role in developing problem-specific and quantum-aware models capable of achieving a computational advantage. Despite the simplicity of the main idea of GQML -- create architectures respecting the symmetries of the data -- its practical implementation requires a significant amount of knowledge of group representation theory. We present an introduction to representation theory tools from the optics of quantum learning, driven by key examples involving discrete and continuous groups. These examples are sewn together by an exposition outlining the formal capture of GQML symmetries via "label invariance under the action of a group representation", a brief (but rigorous) tour through finite and compact Lie group representation theory, a reexamination of ubiquitous tools like Haar integration and twirling, and an overview of some successful strategies for detecting symmetries., Comment: 43 pages, 10 figures. Updated to add relevant references

The reservoir of molecular gas (H$_{\rm 2}$) represents the fuel for the star formation (SF) of a galaxy. Connecting the star formation rate (SFR) to the available H$_{\rm 2}$ is key to accurately model SF in cosmological simulations of galaxy formation. We investigate how modifying the underlying modelling of H$_{\rm 2}$ and the description of stellar feedback in low-metallicity environments (LMF, i.e. low-metallicity stellar feedback) in cosmological, zoomed-in simulations of a Milky Way-size halo influences the formation history of the forming, spiral galaxy and its final properties. We exploit two different models to compute the molecular fraction of cold gas (f$_{\rm H_{\rm 2}}$): $i)$ the theoretical model by Krumholz et al. (2009b) and $ii)$ the phenomenological prescription by Blitz & Rosolowsky (2006). We find that the model adopted to estimate f$_{\rm H_{\rm 2}}$ plays a key role in determining final properties and in shaping the morphology of the galaxy. The clumpier interstellar medium (ISM) and the more complex H$_{\rm 2}$ distribution that the Krumholz et al. (2009b) model predicts result in better agreement with observations of nearby disc galaxies. This shows how crucial it is to link the SFR to the physical properties of the star-forming, molecular ISM. The additional source of energy that LMF supplies in a metal-poor ISM is key in controlling SF at high redshift and in regulating the reservoir of SF across cosmic time. Not only is LMF able to regulate cooling properties of the ISM, but it also reduces the stellar mass of the galaxy bulge. These findings can foster the improvement of the numerical modelling of SF in cosmological simulations., Comment: accepted for publication in MNRAS

We study the evolution of dust in a cosmological volume using a hydrodynamical simulation in which the dust production is coupled with the MUPPI (MUlti Phase Particle Integrator) sub-resolution model of star formation and feedback. As for the latter, we keep as reference the model setup calibrated previously to match the general properties of Milky Way like galaxies in zoom-in simulations. However, we suggest that an increase of the star formation efficiency with the local dust to gas ratio would better reproduce the observed evolution of the cosmic star formation density. Moreover, the paucity of quenched galaxies at low redshift demands a stronger role of AGN feedback. We tune the parameters ruling direct dust production from evolved stars and accretion in the inter stellar medium to get scaling relations involving dust, stellar mass and metallicity in good agreement with observations. In low mass galaxies the accretion process is inefficient. As a consequence, they remain poorer in silicate and small grains than higher mass ones. We reproduce reasonably well the few available data on the radial distribution of dust outside the galactic region, supporting the assumption that the dust and gas dynamics are well coupled at galactic scales., Comment: MNRAS accepted

In this paper, we assess the impact of numerical resolution and of the implementation of energy input from AGN feedback models on the inner structure of cluster sub-haloes in hydrodynamic simulations. We compare several zoom-in re-simulations of a sub-sample of the cluster-sized haloes studied in Meneghetti et al. (2020), obtained by varying mass resolution, softening length and AGN energy feedback scheme. We study the impact of these different setups on the subhalo abundances, their radial distribution, their density and mass profiles and the relation between the maximum circular velocity, which is a proxy for subhalo compactness. Regardless of the adopted numerical resolution and feedback model, subhaloes with masses Msub < 1e11Msun/h, the most relevant mass-range for galaxy-galaxy strong lensing, have maximum circular velocities ~30% smaller than those measured from strong lensing observations of Bergamini et al. (2019). We also find that simulations with less effective AGN energy feedback produce massive subhaloes (Msub> 1e11 Msun/h ) with higher maximum circular velocity and that their Vmax - Msub relation approaches the observed one. However the stellar-mass number count of these objects exceeds the one found in observations and we find that the compactness of these simulated subhaloes is the result of an extremely over-efficient star formation in their cores, also leading to larger-than-observed subhalo stellar mass. We conclude that simulations are unable to simultaneously reproduce the observed stellar masses and compactness (or maximum circular velocities) of cluster galaxies. Thus, the discrepancy between theory and observations that emerged from the analysis of Meneghetti et al. (2020) persists. It remains an open question as to whether such a discrepancy reflects limitations of the current implementation of galaxy formation models or the LCDM paradigm., Comment: 11 pages, 10 figures, abstract is redacted to fit arXiv character count limit

Meneghetti et al. (2020) recently reported an excess of galaxy-galaxy strong lensing (GGSL) in galaxy clusters compared to expectations from the LCDM cosmological model. Theoretical estimates of the GGSL probability are based on the analysis of numerical hydrodynamical simulations in the LCDM cosmology. We quantify the impact of the numerical resolution and AGN feedback scheme adopted in cosmological simulations on the predicted GGSL probability and determine if varying these simulation properties can alleviate the gap with observations. We repeat the analysis of Meneghetti et al. (2020) on cluster-size halos simulated with different mass and force resolutions and implementing several independent AGN feedback schemes. We find that improving the mass resolution by a factor of ten and twenty-five, while using the same galaxy formation model that includes AGN feedback, does not affect the GGSL probability. We find similar results regarding the choice of gravitational softening. On the contrary, adopting an AGN feedback scheme that is less efficient at suppressing gas cooling and star formation leads to an increase in the GGSL probability by a factor between three and six. However, we notice that such simulations form overly massive subhalos whose contribution to the lensing cross-section would be significant while their Einstein radii are too large to be consistent with the observations. The primary contributors to the observed GGSL cross-sections are subhalos with smaller masses, that are compact enough to become critical for lensing. The population with these required characteristics appears to be absent in simulations., Comment: 13 pages, 11 figures. Submitted for publication on Astronomy and Astrophysics